Niels Johannesen and Gabriel Zucman The end of bank ...eprints.lse.ac.uk/56125/1/__lse.ac.uk_storage... · bank secrecy. Exploiting a unique panel dataset, our study is the rst attempt

Niels Johannesen and Gabriel Zucman

The end of bank secrecy? An evaluation of the G20 tax haven crackdown Article (Accepted version) (Refereed)

Original citation: Johannesen, Niels and Zucman, Gabriel (2014) The end of bank secrecy? An evaluation of the G20 tax haven crackdown. American Economic Journal: Economic Policy, 6 (1). pp. 65-91. ISSN 1945-7731 DOI: 10.1257/pol.6.1.65 © 2014 AEA This version available at: http://eprints.lse.ac.uk/56125/ Available in LSE Research Online: March 2014 LSE has developed LSE Research Online so that users may access research output of the School. Copyright © and Moral Rights for the papers on this site are retained by the individual authors and/or other copyright owners. Users may download and/or print one copy of any article(s) in LSE Research Online to facilitate their private study or for non-commercial research. You may not engage in further distribution of the material or use it for any profit-making activities or any commercial gain. You may freely distribute the URL (http://eprints.lse.ac.uk) of the LSE Research Online website. This document is the author’s final accepted version of the journal article. There may be differences between this version and the published version. You are advised to consult the publisher’s version if you wish to cite from it.

http://www.aeaweb.org/aej/pol/index.php

http://dx.doi.org/10.1257/pol.6.1.65

http://www.vanderbilt.edu/AEA/

http://eprints.lse.ac.uk/56125/

The End of Bank Secrecy? An Evaluation of the G20 TaxHaven Crackdown

By Niels Johannesen and Gabriel Zucman∗

During the financial crisis, G20 countries compelled tax havens tosign bilateral treaties providing for exchange of bank information.Policymakers have celebrated this global initiative as the end ofbank secrecy. Exploiting a unique panel dataset, our study is thefirst attempt to assess how the treaties affected bank deposits intax havens. Rather than repatriating funds, our results suggestthat tax evaders shifted deposits to havens not covered by a treatywith their home country. The crackdown thus caused a relocationof deposits at the benefit of the least compliant havens. We discussthe policy implications of these findings.JEL: H26, H87, G21, K42Keywords: Tax havens, Tax evasion

In August 2009, France and Switzerland amended their tax treaty. The twocountries agreed to exchange upon request all information necessary for tax en-forcement, including bank information otherwise protected by Swiss bank secrecylaws. Over the following months, one of France’s richest persons and her wealthmanager were taped discussing what to do with two undeclared Swiss bank ac-counts, worth $160 million. After a visit to Switzerland, the wealth managerconcluded that keeping the funds in Swiss banks or bringing them back to Francewould be too risky. He suggested that the funds be transferred to Hong Kong,Singapore, or Uruguay, three tax havens which had not committed to exchanginginformation with France. After the tapes were made public, they received exten-sive newspaper coverage and eventually the funds were repatriated to France.1

The amendment to the French-Swiss tax treaty was part of a major initiativeto combat tax evasion at the global level. Since the end of the 1990s, the OECDhas encouraged tax havens to exchange information with other countries on thebasis of bilateral tax treaties, but until 2008 most tax havens declined to sign

∗ Johannesen: Department of Economics, University of Copenhagen, Øster Farimagsgade 5, 1353Copenhagen, Denmark, [email protected]. Zucman: Paris School of Economics, 48 boulevardJourdan, 75014 Paris, France, [email protected]. We are grateful to the Bank for International Settle-ments for giving us access to the otherwise restricted data on foreign deposits for the purposes of thisresearch and to the Danish Central Bank for facilitating the contact. We thank two anonymous refer-ees, Nicolas Frmeaux, Lucie Gadenne, Harry Huizinga, Thomas Piketty, Johannes Voget, audiences atthe Danish Central Bank, University of Copenhagen, Paris School of Economics, Max Planck Institute,Norwegian Business School, Oxford, Tilburg, Vienna Institute for Advanced Studies, IIPF Congress,Days Louis-Andre Gerard-Varet, and Helsinki for valuable comments and suggestions. Niels Johannesenacknowledges financial support from the Danish Council for Independent Research and Gabriel Zucmanfrom the French Ministry of Higher Education.

1For a summary of this evasion case, see “Affaire Bettencourt: ce que disent les enregistrements,” LeMonde, 30 June 2010.

1

2 AMERICAN ECONOMIC JOURNAL MONTH YEAR

such treaties. During the financial crisis, the fight against tax evasion became apolitical priority in rich countries and the pressure on tax havens mounted. Atthe summit held in April 2009, G20 countries urged each tax haven to sign atleast 12 information exchange treaties under the threat of economic sanctions.Between the summit and the end of 2009, the world’s tax havens signed a totalof more than 300 treaties.

The effectiveness of this crackdown on offshore tax evasion is highly contested.A positive view asserts that treaties significantly raise the probability of detectingtax evasion and greatly improve tax collection (OECD, 2011). According to policymakers, “the era of bank secrecy is over” (G20, 2009). A negative view, on thecontrary, asserts that the G20 initiative leaves considerable scope for bank secrecyand brings negligible benefits (Shaxson and Christensen, 2011). Whether thepositive or the negative view is closer to reality is the question we attempt toaddress in this paper.

This is an important question for two reasons. First, the fight against offshoretax evasion is a key policy issue. Globalization and the information technologyrevolution have made it easier for tax evaders to move funds offshore. Absentinformation exchange between countries, personal capital income taxes cannot beproperly enforced, giving rise to substantial revenue losses and constraining thedesign of tax systems. Against the backdrop of the large public deficits faced bymost countries since the financial crisis, curbing tax evasion is high on the policyagenda.

Second, although treaties have prevailed as the main policy instrument in thefight against international tax evasion, surprisingly little is known about their ef-fectiveness. The G20 crackdown has generated a lot of discussion in policy circlesbut there is little fact-based evidence of its efficacy and no academic evaluation.The OECD has launched a peer-review evaluation to assess whether treaties areproperly drafted and enforced, but while this legal work is necessary, it is notsufficient: if the information exchange mechanism advocated by the OECD hasfundamental shortcomings, then even properly drafted and enforced treaties maybe ineffective. Our study is the first attempt to assess from a quantitative per-spective the impact of the many treaties signed by tax havens since G20 countrieshave made tax evasion a priority.

Providing compelling evidence on tax evasion is notoriously difficult, and evenharder in the complex area of international tax evasion. We break new ground inthis field by drawing on a particularly rich dataset on cross-border bank deposits.For the purpose of our study, the Bank for International Settlements (BIS) hasgiven us access to bilateral bank deposit data for 13 major tax havens, includingSwitzerland, Luxembourg, and the Cayman Islands. We thus observe the valueof the deposits held by French residents in Switzerland, by German residents inLuxembourg, by U.S. residents in the Cayman Islands and so forth, on a quarterlybasis from the end of 2003 to the middle of 2011. Using specific country names forthe sake of concreteness, we ask: Did French holders of Swiss deposits respond to

VOL. VOL NO. ISSUE THE END OF BANK SECRECY? 3

the 2009 French-Swiss treaty by repatriating funds to France? Did they relocatetheir funds to other tax havens? Or did they simply leave them in Switzerland?To address these questions, after providing more details on offshore tax evasionand the data we use in Section I, we employ graphical analysis in Section II andpanel regression analysis in Section III.

We obtain two main results. First, treaties have had a statistically significantbut quite modest impact on bank deposits in tax havens: a treaty between sayFrance and Switzerland causes an approximately 11 percent decline in the Swissdeposits held by French residents. Second, and more importantly, the treatiessigned by tax havens have not triggered significant repatriations of funds, butrather a relocation of deposits between tax havens. We observe this pattern inthe aggregate data: the global value of deposits in havens remains the same twoyears after the start of the crackdown, but the havens that have signed manytreaties have lost deposits at the expense of those that have signed few. Wealso observe this pattern in the bilateral panel regressions: after say France andSwitzerland sign a treaty, French deposits increase in havens that have no treatywith France.

The finding that tax evaders shift deposits in response to treaties, our key result,illustrates an important pitfall of the current approach to the fight against taxevasion. Tax havens are whitelisted after signing 12 treaties, leaving considerablescope for tax evaders to ensure that their assets are not covered by a treaty. Ouranalysis shows that tax evaders exploit this possibility, which ultimately providesincentives for tax havens to keep their treaty networks at the minimum. From anormative viewpoint, our paper thus lends support to the idea developed theoret-ically by Elsayyad and Konrad (2011) that a “big bang” multilateral agreementshould be preferred to the current sequential approach.

The finding that treaties have had a modestly sized impact on bank depositshas several possible interpretations between which we cannot discriminate con-clusively with the data at our disposal. First, most tax evaders may have chosennot to move deposits because they considered that treaties did not substantiallyincrease the probability they be detected. This interpretation is consistent withthe fact that treaties only rarely lead to actual exchange of information in prac-tice. Yet another possible interpretation is that the modest size of our estimatesis due to limitations of our deposit dataset. For instance, some tax evaders usesham corporations with addresses in Panama and the British Virgin Islands asnominal holders of their bank accounts in Switzerland and other havens, whichobscures who ultimately owns part of the funds offshore. We tackle this issuein Section IV, for the first time in this literature, and we show that the fundsheld through sham corporations might have responded strongly to the treaties.Lastly, tax evaders might have declared some of their assets to tax authoritieswhile keeping them offshore. In Section V we analyze a novel dataset with di-rect information on income that European owners of Swiss accounts voluntarilydeclare. We find no signs that treaties induced Swiss account holders to comply


more with tax laws, but we cannot rule out an increase in compliance in othertax havens.

Our paper adds to the literature on tax treaties, where a recurring finding isthat treaties have little real economic effects (e.g., Blonigen and Davies, 2005;di Giovanni, 2005; Louie and Rousslang, 2008). Relative to this literature, ourcontribution is to focus on the information sharing provisions included in taxtreaties rather than on those aimed at promoting cross-border investments andlimiting double taxation. The effectiveness of information sharing mechanisms israrely assessed and our paper contributes to filling this gap.2

We also contribute to the literature on how tax policies affect internationalinvestments (e.g., Chan, Covrig and Ng, 2005; Desai and Dharmapala, 2011). Abranch of this literature initiated by Alworth and Andresen (1992) focuses onthe determinants of cross-border deposits such as taxes, interest rate differentialsand distance. Huizinga and Nicodeme (2004) find that information exchangeagreements have no significant effect on cross-border deposits in OECD countries.We focus, by contrast, on how tax treaties affect deposits in tax havens. Thisevaluation was not possible before 2009, the year when most tax havens startedsigning information exchange treaties.3

Lastly, our paper sheds new light on the activities taking place in tax havens,a topic which is attracting increasing interest (Desai, Foley and Hines, 2006;Dharmapala, 2008; Dharmapala and Hines, 2009; Palan, Murphy and Chav-agneux, 2010). Tax havens provide corporations and individuals with oppor-tunities to avoid or evade taxes. The bulk of the literature focuses on the useof tax havens by corporations, following Hines and Rice (1994). By contrast, wefocus on their use by households, which is still little studied.

I. Offshore Tax Evasion By Households: Context and Data

A. Policies to Prevent Offshore Tax Evasion

Tax havens such as Switzerland, Singapore, and the Cayman Islands host animportant wealth management industry which provides foreigners with an oppor-tunity to evade taxes. If a French household entrusts assets to a French bank,there is automatic reporting of capital income to the French tax authorities: eva-sion of the personal income tax is impossible. But if it entrusts assets to a Swissbank, there is no automatic reporting: French authorities have to rely on self-reporting and tax evasion is possible.4 Using official Swiss statistics and anomaliesin the international investment data of countries, Zucman (2013) estimates that

2A complementary contribution is Blonigen, Oldenski and Sly (2011) who study whether informationexchange agreements affect foreign direct investments (while we look at bank deposits and tax evasion).

3Two related papers are Hemmelgarn and Nicodeme (2009) and Johannesen (2010), who study theeffects of the Savings Directive, a European policy initiative that imposes a tax on interest income earnedby European Union residents in a number of tax havens. We discuss in the conclusion the relative meritsof withholding taxes and treaties in light of our results.

4 Kleven et al. (2011) document the importance of third-party reporting to prevent tax evasion.


around 8 percent of households’ global financial wealth is held in tax havens. Thisfigure implies substantial tax revenue losses due to outright fraud.

Missing information on income earned through bank accounts in tax havensis the key problem for enforcing personal capital income taxes. Exchange of in-formation between countries is the obvious solution. There are two main wayscountries can exchange information: automatically or upon request (Keen andLigthart, 2006). Automatic exchange of information is widely acknowledged tobe the most effective solution because it allows tax authorities to obtain com-prehensive data about income earned by domestic residents in foreign banks.But information exchange upon request is more common. It is the standardpromoted by the OECD and embedded in the treaties signed by tax havens. Un-der the amended French-Swiss treaty, French authorities can request informationfrom Switzerland to enforce tax laws. Requests must concern specific taxpayers.France cannot ask for a list of all its residents with funds in Switzerland. More-over, the requested information must be “foreseeably relevant” (OECD, 2008, p.38): information can be obtained by French authorities only if they have a welldocumented suspicion that a resident is evading taxes. All the treaties signed bytax havens have identical wording: they follow the OECD model tax convention.

The usefulness of the OECD standard of information exchange is the objectof much controversy. Critics argue that since placing a request for informationrequires prior knowledge, which is extremely hard to come about, little can beobtained through treaties (Sheppard, 2009). And indeed, the U.S. GovernmentAccountability Office (2011) revealed that during the 2006-2010 period, the U.S.placed only 894 requests under its more than 80 tax treaties. Since a single Swissbank admitted in 2008 to have more than 19,000 U.S. clients with undeclaredbank accounts (U.S. Senate, 2008), information exchange upon request is clearlyassociated with a small probability of detecting tax evasion. Advocates of theOECD standard, on the other hand, stress that even a small probability of detec-tion may be sufficient to deter tax evasion and that information exchange uponrequest is a major step forward from no exchange at all.

Since the end of the 1990s, the OECD has tried to convince tax havens tosign information exchange treaties. But, as shown by Figure 1, most havensdeclined to sign treaties until the financial crisis.5 The turning point occurred inApril 2009. The OECD specified that each tax haven should conclude at least 12treaties to be in compliance and drew up a list of 42 non-compliant havens. TheG20 threatened to impose economic sanctions on non-compliant havens. In justfive days, all havens committed to signing 12 treaties and the G20 declared theera of bank secrecy over (G20, 2009).

As a result of G20 pressure, treaty signature effectively boomed in 2009 and2010. But the pace slowed down considerably after 2010. Moreover, tax havenssigned many treaties with each other: in 2009, almost one-third of the treaties

5All the data on tax treaties and aggregate bank deposits used for this research are available onlineon the authors’ websites.


0

100

200

300

400

2004 2005 2006 2007 2008 2009 2010 2011S1

Treaties with other tax havens

Treaties with non-haven countries

Figure 1. Number of Bank Information Exchange Treaties Signed by Tax Havens, by Year

Note: The figure charts the the number of new treaties or amendment to existing treaties allowing forinformation exchange signed each year by the world’s 52 tax havens (see list in the Online Appendix).Source: Global Forum on Transparency and Exchange of Information for Tax Purposes (1998-2011)and authors’ research (see Online Appendix).

signed by tax havens were with other havens. Such haven-haven treaties do nothelp non-haven countries curb tax evasion in any way. In all likelihood they onlyreflect the desire of some tax havens to reach the 12 treaties threshold withoutgiving substantial concessions.

B. Data on Tax Treaties

To study the effects of the G20 tax haven crackdown, we have compiled acomplete dataset on the treaties concluded by tax havens. The dataset covers 52tax havens (see the Online Appendix), more than 220 potential partner countries,and includes information until the end of 2011q2.

Tax havens can start exchanging information with partner countries on thebasis of two types of legal events: new treaties or amendments to existing treatieson the one hand (for instance, the amendment to the French-Swiss tax treaty inAugust 2009), and changes in domestic laws allowing for information exchangewith existing treaty partners on the other (Cyprus passed such a law in July2008). The two types of events are legally equivalent, but new treaties may bemore salient than subtle changes in the banking laws of tax havens. Distinguishingbetween the two kinds of legal events allows us to investigate whether depositors


respond differently to more salient events.6

The main data source is the Exchange of Tax Information Portal, which repre-sents the best effort of the OECD to gather accurate information on tax treaties.7

In some cases, we have added information from official government websites. TheOnline Appendix describes step-by-step how we compiled the treaty dataset fromreadily available sources. The final dataset includes 1,025 events: 861 new treatiesor amendments to existing treaties, and 164 instances when changes in domesticlaws rendered information exchange possible under existing treaties. Note thatsince there are 52 tax havens and around 220 countries and territories in theworld, a full network of treaties would include around 11,000 treaties. Througha peer-review evaluation, the OECD assesses whether the treaties signed by taxhavens are properly drafted and enforced. Out of the 861 new treaties signed from2004 to mid-2011, 68 percent were deemed compliant, 13 percent were deemednot compliant, and 19 percent were still unreviewed in November 2011.

C. Data on Deposits in Tax Havens

Our second data source is the BIS locational banking statistics, which containinformation on foreign bank deposits in 41 countries. The BIS publishes quarterlydata aggregated at the country level, for instance total deposits held by Frenchresidents in foreign banks and total deposits held by foreign residents in Swissbanks. For our study and on the condition that we do not disclose bilateralinformation, the BIS has given us access to deposit data at the bilateral level,for instance deposits held by French residents in Swiss banks. There are 18 taxhavens reporting to the BIS. We have access to bilateral deposit data for 13 ofthem: Austria, Belgium, the Cayman Islands, Chile, Cyprus, Guernsey, the Isleof Man, Jersey, Luxembourg, Macao, Malaysia, Panama, and Switzerland. Wealso have bilateral data for the aggregate of the remaining 5 havens: Bahamas,Bahrain, Hong Kong, the Netherlands Antilles, and Singapore.8 The 13 havensfor which we have bilateral data host about 75 percent of the deposits of all BIS-reporting havens, which allows us to make reasonable inference from this sampleof countries.

The BIS locational banking statistics are widely used in international economicsand are a key input to statistics on balance of payments. The most importantfinancial centers (havens and non-havens) report to the BIS. New financial centersare systematically included in the BIS statistics once they reach a significant size,so that the havens not covered are by construction very small. Further, within

6 Chetty, Looney and Kroft (2009) provide evidence of the importance of salience for the response totaxes.

7See http://eoi-tax.org/. We have also benefited from discussions with Jeremy Maddison andSanjeev Sharma from the OECD.

8The secession of the Netherlands Antilles in October 2010 resulted in two new countries, Curaao andSint Maarten. Curaao took over the reporting obligation to the BIS. Note also that we do not includeBermuda in our list of tax havens, because there are no private wealth management activities there (only4 banks are registered in Bermuda).

http://eoi-tax.org/


each covered center there is almost full coverage of deposits, because all the bankswith cross-border positions in excess of a modest threshold (e.g., $10 million in theBahamas) are required to report. The Bank for International Settlements (2006)indicates that coverage rates systematically exceed 90 percent. The reportingrequirements of the BIS do not violate any bank secrecy provisions, because banksdo not report data on individual customers but only aggregate figures.

The BIS data, however, have three limitations. First, it is not possible to knowwhat fraction of the deposits in tax havens belong to households evading taxes.The BIS provides a sectoral decomposition between deposits owned by banks andby “non-banks.” Since interbank deposits do not play a role in personal incometax evasion, we focus on the deposits of “non-banks.” Part of these deposits,however, belong to multinational corporations that stash cash offshore and thatare not affected by bank information sharing. Ideally we would like to observethe deposits that belong to households only. Since this is not possible, we cannotdirectly estimate the behavioral response of tax evaders: all we can do is makinginference from the evolution of the deposits owned by “non-banks.”

To do so, we need an idea of what fraction of “non-bank” deposits belong tohouseholds. Data made available by a number of BIS-participating central banksenable us to shed light on this issue. In Switzerland, the second largest offshorecenter in terms of “non-bank” deposits, 80-90 percent of the deposits seem tobelong to households.9 The Bank of England reports that in 2007 householdsowned about 70-75 percent of the deposits in the Channel Islands and the Isle ofMan, collectively the third largest offshore center. And a previous study (Zucman,2013), using different data, found that at least 50 percent of haven deposits likelybelong to households.10 On the basis of these elements, our baseline assumptionwhen we interpret the results will be that tax evaders own about 50 percent ofthe deposits in tax havens.

The second limitation of the BIS data is that they are based on immediaterather than beneficial ownership. If a French individual owns a Swiss depositthrough a sham corporation with an address in Panama, the BIS assigns thefunds to Panama. Almost 25 percent of all deposits in tax havens are registeredas belonging to other havens reflecting the widespread use of sham corporationsby clients of offshore banks. Our analysis in Section IV will explicitly address the

9There are two types of Swiss bank deposits covered by the BIS data: regular deposits (10-20 percentof the total) and “fiduciary deposits” (80-90 percent). In all likelihood, fiduciary deposits entirely belongto individuals: these are investments made by Swiss banks in foreign money markets on behalf of foreignindividuals, an arrangement that enables clients of Swiss banks to avoid the 35 percent tax imposed bySwitzerland on Swiss-source capital income. Multinational corporations do not use fiduciary depositsbecause they can directly invest in foreign money markets without having to pay the handsome feescharged by Swiss banks for these operations. For more details on fiduciary deposits, see e.g. Brown,Dobeli and Saure (2011).

10The figure was obtained as follows. On the basis of official Swiss National Bank statistics and of largeanomalies in the international investment data of countries, Zucman (2013) estimates that individualsowned at least $6tr in financial assets through bank accounts in tax havens in end 2008, of which $1.4 trtook the form of bank deposits. These $1.4 tr account for 50 percent of the total deposits in tax havensas per the BIS.


existence of deposits held through sham corporations.

0

1,000

2,000

3,000

4,000

5,000

0

500

1,000

1,500

2,000

2,500

3,000

3,500

2004 2005 2006 2007 2008 2009 2010 2011

Cross-border deposits in tax havens (left-hand scale)

Cross-border deposits in non-haven countries (right-hand scale)

Figure 2. Bank Deposits in Haven and Non-Haven Countries, 2004-2011 (bn US$)

Note: All figures are yearly averages (first semester-average for 2011).

Source: Bank for International Settlements (2002-2011, Table 3B).

Lastly, the BIS data relate to only one form of wealth held by households intax havens: bank deposits. They do not provide information on the equity andbond portfolios that savers entrust to tax haven banks. There is little publicinformation on households’ offshore portfolios, except in Switzerland. The SwissNational Bank reports that about 25 percent of the funds held by foreigners inSwitzerland take the form of bank deposits, while 75 percent are equities andbonds (Zucman, 2013). With the data at our disposal, we cannot say anythingabout the response of tax evaders’ portfolio wealth to treaties: we can only analyzethe evolution of deposits. It is safe, however, to assume that the response of bankdeposits is a good proxy for the response of the overall stock of offshore wealth,because the information exchange provisions of treaties affect all assets similarly.

II. Graphical Evidence

A. The Effects of the G20 Initiative on Aggregate Deposits

As a starting point for the empirical analysis, Figure 2 shows the evolution ofthe bank deposits held on aggregate in the 18 tax havens reporting to the BIS.


Despite the wave of treaties signed in 2009-2010, deposits in tax havens remainedstable over the 2007-2011 period at around $2,700 billion. For comparison, thefigure shows the evolution of the deposits held on aggregate in the non-havencountries reporting to the BIS. This group includes financial centers that have alarge treaty network and have not been affected by the G20 initiative, such asthe U.S. or Germany. Deposits in havens and non-havens have followed a similartrend over the 2004-2011 period. The evolution of deposits in non-havens mightbe an imperfect counterfactual for the evolution of deposits in tax havens, but wecan at least exclude that the G20 crackdown was followed by a significant dropin aggregate deposits in tax havens.

0

200

400

600

800

1,000

1,200

0

100

200

300

400

500

600

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Country-pairs that signed a compliant treaty between 2008q1 and 2011q2, left-hand scale

Country-pairs that did not sign a compliant treaty between 2008q1 and 2011q2, right-hand scale

Figure 3. Bank Deposits in Treaty and No-Treaty Country-Pairs, 2002-2011 (bn US$)

Note: The figure charts the evolution of the deposits held by savers of country i in banks of tax havenj for the set of country-haven pairs (i, j) that signed a treaty deemed compliant by the OECD betweenJanuary 1st 2008 and June 30th 2011, and the set of country-haven pairs that did not. Saver countriesexclude tax havens. Tax havens include Austria, Belgium, Chile, the Cayman Islands, Cyprus, Guernsey,the Isle of Man, Jersey, Luxembourg, Macao, Malaysia, Panama, and Switzerland. All figures are yearlyaverages (first semester-average for 2011) and expressed in billion U.S. dollars.Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.

Next, we compare the deposits that have become covered by a treaty to thedeposits that have not. We consider all country-haven combinations (e.g., France-Switzerland) among the 13 havens for which we have bilateral deposit data andthe more than 200 countries holding deposits in these havens. From this universe,we construct two groups: a “treaty” group including all country-haven pairs thatsigned a compliant treaty between 1 January 2008 and 30 June 2011, and a “no-


treaty” group including all other pairs. Figure 3 shows that deposits decreasedmoderately in the “treaty” group but remained roughly stable in the “no-treaty”group. Should all deposits have followed the same trend, the deposits in the“treaty” group would have been around 15 percent larger in 2011. Figure 3suggests that at least some tax evaders responded to treaty signatures, althoughit does not reveal the nature of this response.

B. The Effects of the G20 Initiative on the Deposits in Each Tax Haven

To investigate how tax evaders responded to treaties, we examine the evolutionof deposits in each tax haven between 2007 and 2011. Figure 4 reveals that theglobally stable level of deposits in tax havens conceals significant differences acrosshavens. Banks in Jersey lost the equivalent of 4 percent of the 2007 total amountof haven deposits (i.e., about 8 percent of tax evaders’ deposits, if tax evaders ownabout 50 percent of haven deposits), while banks in Hong Kong gained around2.5 percent (about 5 percent of tax evaders’ deposits).

Jers

ey

Luxe

mbo

urg

Sw

itzer

land

Isle

of M

an

Gue

rnse

y

Bah

rain

Net

herla

nds

Ant

illes

Ber

mud

a

Bah

amas

Mac

ao

Mal

aysi

a

Chi

le

Pan

ama

Cyp

rus

Aus

tria

Bel

gium

Sin

gapo

re

Cay

man

Hon

g K

ong

All

tax

have

ns

-5%

-4%

-3%

-2%

-1%

0%

1%

2%

3%

4%

5%

Evolution of bank deposits in tax havens between 2007 and 2011, as a percentage of total 2007 deposits in tax havens

Figure 4. Evolution of Bank Deposits in Each Tax Haven, 2007-2011 (bn US$)

Note: The figure charts the evolution of the foreign-owned deposits in each BIS-reporting tax haven. Wecompare first semester of 2011 averages with 2007 averages (except for Cyprus which started reportingin 2008q4 and Malaysia which started in 2007q4), and express the difference as a fraction of the depositsheld in all tax havens in 2007 ($2,600bn).Source: Bank for International Settlements (2002-2011, Table 3B).

Crucially, the deposit gains and losses correlate strongly with the number oftreaties signed by each haven. Figure 5 plots the percentage change of each haven’s


deposits between 2007 and 2011 against the number of compliant treaties signedover the same period. Cyprus signed only 2 compliant treaties and experienced a60 percent increase in its deposits, whereas Guernsey signed 19 compliant treatiesand experienced a 15 percent decrease. A simple bivariate regression suggests thatan additional treaty signed by a haven is associated with a decrease of 3.8 percentof the deposits in its banks (with a standard error of 1.4 percent).11

Austria

Bahamas

Bahrain

Belgium Cayman Islands

Cyprus

Guernsey

Hong-Kong

Isle of Man

Jersey

Luxembourg

Macao

Malaysia

Netherlands Antilles

Panama

Singapore

Switzerland

b = - 0.038 (0.014)

-80%

-60%

-40%

-20%

0%

20%

40%

60%

80%

0 5 10 15 20 25

Gro

wth

rat

e of

dep

osits

bet

wee

n 20

07 a

nd 2

011

S1

Number of compliant treaties signed with non-haven countries between January 2008 and June 2011

Figure 5. Deposit Growth and Treaty Signature Activity of Tax Havens, 2007-2011

Note: The figure charts the growth rate of the deposits in each BIS-reporting tax haven between 2007(year average, except for Cyprus which started reporting in 2008q4 and Malaysia which started in 2007q4)and 2011 (first semester average), as a function of the number of compliant treaties signed between thebeginning of 2008 and the end of the first semester 2011. b is the coefficient of the slope with standarderror in parentheses.Source: Deposits: Bank for International Settlements (2002-2011, Table 3B). Compliant treaties: GlobalForum on Transparency and Exchange of Information for Tax Purposes (1998-2011) and authors’ research,see Online Appendix.

Overall, the graphical evidence suggests that a number of tax evaders respondedto treaties and that their response was mostly to transfer deposits to other taxhavens, leaving roughly unchanged the funds globally held in tax havens. Figure6 lends additional support to this conjecture. It shows that there is no correla-tion between the number of treaties signed by OECD countries with tax havensbetween 2007 and 2011 and the growth of the deposits held by OECD countries’residents in tax havens. Signing more treaties does not seem to help OECD

11This correlation remains when we consider cumulated exchange rate adjusted net flows in each havenas a percentage of end-2007 stocks rather than the simple growth rate of deposits, or when we considerall treaties signed, whether complying with the OECD standard, unreviewed, or not complying.


countries repatriate funds.

Australia

Austria

Belgium Canada

Chile Czech Republic

Denmark Estonia

Finland

France

Germany

Greece

Hungary

Iceland

Ireland

Israel Italy

Japan

Luxembourg

Mexico

Netherlands New Zealand Norway

Poland Portugal

Slovakia Slovenia

South Korea

Spain

Sweden

Switzerland

Turkey UK

USA

b = - 0.006 (0.02)

-100%

-50%

0%

50%

100%

150%

200%

250%

0 5 10 15

Gro

wth

rat

e of

dep

osits

in B

IS-r

epor

ting

have

ns

betw

een

2007

and

201

1 S

1

Number of compliant treaties signed with BIS-reporting havens between January 2008 and June 2011

Figure 6. Deposit Growth and Treaty Signature Activity of OECD countries, 2007-2011

Note: The figure charts the growth rate of the deposits held by each OECD country in BIS-reportingtax havens between 2007 (year average) and 2011 (first semester average), as a function of the numberof compliant treaties signed between the beginning of 2008 and the end of the first semester 2011. b isthe coefficient of the slope with standard error in parentheses.Source: Deposits: Bank for International Settlements (2002-2011), restricted bilateral locational bankingstatistics. Compliant treaties: Global Forum on Transparency and Exchange of Information for TaxPurposes (1998-2011) and authors’ research, see Online Appendix.

While the graphical evidence suggests a consistent scenario, it aggregates treatiessigned at different dates and does not fully exploit the bilateral nature of our data.To deal with this, we now turn to panel regression analysis.

III. Regression-Based Evidence

A. The Impact of Treaties on Bilateral Deposits

The first question we want to address is whether treaties have had a statisti-cally significant impact on deposits in tax havens at the bilateral level. We runregressions of the form:

(1) log(Depositsijq) = α+ βSignedijq + γij + θq + εijq


where Depositsijq denotes the deposits held by residents of country i with banksof haven j at the end of quarter q, Signedijq is a dummy equal to 1 if a treatyallowing for information exchange between i and j exists in quarter q, γij denotescountry-pair fixed effects, and θq time fixed effects. The coefficient of interest is β:should treaties have any effect at all, β should be statistically different from zero.The country-pair fixed effects γij control for all time invariant characteristics ofcountry-haven pairs, such as distance or common language. The time fixed effectsθq control for all common time trends affecting the deposits in tax havens, suchas the financial crisis. Thus, β only captures the deposit changes in the “treaty”country-haven pairs that come in addition to the deposit changes in the “no-treaty” pairs. All the regressions use the sample period 2003q4-2011q2 and haverobust standard errors clustered at the country-pair level.

The first column of Table 1 estimates equation 1 using the complete universeof country-haven pairs for which we have bilateral deposit data. We find that thedeposits of the “treaty” pairs are smaller after treaty signature than before relativeto the deposits of the “no treaty” pairs. But the coefficient is only borderlinesignificant.

We then in col. (2) restrict the sample to the universe of pairs that include onehaven and one non-haven country, in order for our coefficient β to exclude theeffect of the treaties signed by havens with each other on haven-haven deposits.Treaties now have a larger effect; β is different from zero at the 5 percent level.Col. (3) investigates the effect of haven-haven treaties on haven-haven deposits.We find that a treaty between say the British Virgin Islands (BVI) and Jerseydoes not affect the deposits “held by” the BVI in Jersey, consistent with ournotion that treaties between two havens have no economic meaning. We continuethe analysis with the sample that excludes haven-haven pairs. We refer the readerto Section IV for a detailed analysis of how haven-haven deposits have respondedto treaties between haven and non-haven countries.

In col. (4), we investigate whether depositors respond differently to new treatiesand to changes in the domestic laws of tax havens. Since new treaties are moresalient to tax evaders, we conjecture that evaders should respond more to newtreaties. We interact the dummy variable Signed with dummy variables indicat-ing whether the legal event establishing information exchange is a new treaty ora change in domestic law. The results show that new treaties affect deposits butequivalent changes in domestic laws do not.

The timing of the response to treaty signature is analyzed in col. (5). Weinclude a dummy equal to one in the quarter q of the legal event establishinginformation exchange (Contemp), three dummies equal to one in q+1, q+2, andq + 3 respectively, and a dummy equal to one in all quarters after q + 3. We findthat the bulk of the response occurs two quarters and more after treaty signature.A plausible explanation is that treaties do not enter into force immediately afterthey are signed. For instance, the amendment to the French-Swiss treaty signedin August 2009 entered into force in November 2010. Typically, there is a time


Table 1—Baseline Panel Regressions of Bilateral Bank Deposits on Treaty Signature

(1) (2) (3) (4) (5)BANK: havens BANK: havens BANK: havens BANK: havens BANK: havens

VARIABLES SAVER: all SAVER: non-havens SAVER: havens SAVER: non-havens SAVER: non-havens

Signed -0.0849* -0.1156** 0.0457(0.0893) (0.0349) (0.6926)

Signed × NewTreaty -0.1349**(0.0243)

Signed × DomLaw 0.0163(0.8825)

Signed (Contemp) 0.0223(0.6331)

Signed (+1 quarter) -0.0927(0.1300)

Signed (+2 quarters) -0.1306**(0.0449)

Signed (+3 quarters) -0.1724***(0.0057)

Signed (>3 quarters) -0.1818**(0.0137)

Constant 3.4685*** 3.2187*** 4.3499*** 3.2171*** 3.2196***(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

Observations 39,758 30,960 8,798 30,960 30,960R-squared 0.0870 0.0796 0.1167 0.0798 0.0803Number of panelid 1,631 1,285 346 1,285 1,285Countrypair FE YES YES YES YES YESTime FE YES YES YES YES YES

Note: p-values in parentheses, based on robust standard errors clustered at the country-pair level. ***denotes significance at the 1 percent threshold, ** at the 5 percent threshold, and * at the 10 percentthreshold. The dependent variable is the stock of deposits held by savers of country i in banks of taxhaven j at the end of quarter q. The unit of observation is the country-haven pair (i, j) and the sampleperiod goes from 2003q4 to 2011q2. For a given haven j there are up to 220 saving countries i, and weconsider the deposits held in 13 tax havens j. Signed is a dummy equal to 1 if there exists a treatyproviding for information exchange between i and j in quarter q. NewTreaty is a dummy equal to 1if the event establishing information exchange is a new treaty; DomLaw is a dummy equal to 1 if theevent establishing information exchange is a change in haven’s j domestic law. Signed (Contemp) is adummy equal to 1 in the quarter q when the legal event establishing information exchange between i andj occurs; Signed (+1 quarter) is a dummy equal to 1 in q + 1, and so on.Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.

lag of 3-5 quarters between treaty signature and entry into force.Table 1 confirms that there is a correlation between treaties and deposits in

tax havens: on average, the deposits in the “treaty” pairs decrease after treatysignature relative to the deposits in the “no treaty” pairs. The difference isstatistically significant. But it is quite modest about 11 percent according to col.(2).12 How should we interpret this result?

Because the BIS data include deposits owned by corporations that are not con-cerned by information sharing agreements, our estimated β only provides a lowerbound for the response of tax evaders. If tax evaders own a fraction s of deposits,one can show that their response to treaties is approximately β/s.13 To interpret

12exp(−0.1156) − 1 = 0.10913In a simple difference-in-differences setting in which deposits in the treaty group grow at rate gt and

deposits in the no-treaty group grow at rate gc, the estimator of the response of bank deposits to treaty


what a 11 percent drop in deposits means, we need to take a stance on how larges is. If, as available evidence suggests, s is around 50 percent, then treaties areassociated with a roughly 22 percent average drop in tax evaders’ deposits. Thisis probably much more than expected by those who considered treaties worthless:upon request information exchange seems enough to substantially affect behavior.But it does not seem strong enough to affect the deposit behavior of the majorityof individuals: as long as s is larger than 20-25 percent, our results imply thatonly a minority of tax evaders (weighted by assets) have moved funds in responseto treaties.

Another issue in the interpretation of the magnitude of β is that if tax evadersrespond to treaties by shifting deposits, then our comparison group of “no treaty”country-pairs is also affected by treaty signature. We now augment the model totackle this issue.

B. Deposit Shifting

Table 2 explicitly models shifting behavior. To fix ideas, consider the France-Cayman Islands pair. To explain the amount of French deposits held in theCayman Islands, we introduce in col. (1)-(3) a treaty coverage variable that simplycounts the number of treaties signed by France with the world’s 51 tax havensother than the Cayman Islands. Col. (1) shows that an additional treaty signedby France, say with Switzerland, increases the deposits held by French residents inthe Cayman Islands by 0.6 percent. More generally, it increases French depositsby an average of 0.6 percent in each of the 12 havens other than Switzerlandfor which we have bilateral data. It is natural to assume that deposits are alsoshifted to the havens for which we have no bilateral data, which host around 25percent of offshore deposits. If each haven attracts funds in proportion to itsinitial deposit stock, a treaty signed by France with Switzerland increases Frenchdeposits in each of the world’s havens other than Switzerland by 0.6 percent.14

As col. (2) shows, this shifting only occurs to the benefit of the havens thatdo not have a treaty with France (i.e., when Signed = 0). In such havens, anadditional treaty signed by France is associated with 1.2 percent more French-owned deposits. By contrast, the havens that have a treaty with France (i.e., whenSigned = 1) do not attract deposits. Note also that when we account for shifting,the signature of a treaty between say France and Switzerland still significantlydecreases French deposits in Switzerland, just as we found previously.15

signature (in a log specification) is log[(1 + gt)/(1 + gc)]. If a fraction s of deposits initially belong totax evaders, then the diff-in-diff estimator for the response of tax evaders is: log[(s + gt)/(s + gc)]. At afirst order approximation this is 1/s times larger than log[(1 + gt)/(1 + gc)].

14The fact we do not have bilateral data for all the world’s tax havens does not bias our estimate ofthe magnitude of shifting. Having more bilateral data would simply make our estimate more precise.

15In col. (2) of Table 2, Signed appears in three places, all of which need to be accounted for whencomputing the total effect of an additional treaty on bilateral deposit. Assuming that treaty coverage=6(which is the mean number of compliant treaties signed by OECD countries with tax havens in the 2008-2011 period), the total coefficient on Signed is −0.0498 + 6× (0.0001− 0.0120) = −0.12. This coefficientis comparable to the coefficient found in col. (2) of Table 1.

VO

L.V

OL

NO

.IS

SU

ET

HE

EN

DO

FBA

NK

SEC

REC

Y?

17

Table 2—Panel Regressions of Bilateral Bank Deposits Taking Into Account Deposit Shifting

(1) (2) (3) (4) (5) (6)BANK: havens BANK: havens BANK: havens BANK: havens BANK: havens BANK: havens

SAVER: non-havens SAVER: non-havens SAVER: non-havens SAVER: non-havens SAVER: non-havens SAVER: non-havens

VARIABLESTREATY COVERAGE:

numberTREATY COVERAGE:



shareTREATY COVERAGE:


share

Signed -0.1659*** -0.0498 -0.0750 -0.1468** -0.0816 -0.0933(0.0052) (0.4286) (0.2410) (0.0139) (0.2444) (0.1852)

Saving tax directive (STD) -0.2161*** -0.2198*** -0.1553*** -0.2130*** -0.2135*** -0.1815***(0.0004) (0.0003) (0.0077) (0.0005) (0.0005) (0.0018)

Treaty coverage 0.0059** 0.1272*(0.0402) (0.0568)

Treaty coverage × Signed 0.0001 0.0277(0.9719) (0.7373)

Treaty coverage × (1- Signed) 0.0120*** 0.1752**(0.0033) (0.0318)

Treaty coverage × STD × Signed -0.0030 -0.0679(0.3202) (0.4762)

Treaty coverage × (1-STD) × Signed 0.0066 -0.0927(0.1937) (0.4975)

Treaty coverage × STD × (1-Signed) -0.0071 0.1913*(0.3697) (0.0962)

Treaty coverage × (1-STD) × (1-Signed) 0.0183*** 0.2868***(0.0000) (0.0027)

Constant 3.2147*** 3.2115*** 3.2094*** 3.2285*** 3.2275*** 3.2259***(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

Observations 30,960 30,960 30,960 30,610 30,610 30,610R-squared 0.0829 0.0841 0.0867 0.0835 0.0838 0.0855Number of panelid 1,285 1,285 1,285 1,264 1,264 1,264Countrypair fixed effects YES YES YES YES YES YESTime fixed effects YES YES YES YES YES YES

Note: p-values in parentheses, based on robust standard errors clustered at the country-pair level. *** denotes significance at the 1 percent threshold,** at the 5 percent threshold, and * at the 10 percent threshold. The dependent variable is the stock of deposits held by savers of country i in banks oftax haven j at the end of quarter q. The unit of observation is the country-haven pair (i, j) and the sample period goes from 2003q4 to 2011q2. Signedis a dummy equal to 1 if there exists a treaty providing for information exchange between i and j in quarter q. STD is a dummy equal to one if thecountry-haven pair (i, j) applies the EU Savings Directive. In col. (1)-(3), Treaty coverage counts the number of treaties that i has with tax havens otherthan j. In col. (4)-(6), Treaty coverage measures the share of the deposits held in 2004 by residents of country i in BIS-reporting havens that are coveredby a treaty in quarter q.Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.


Since 2005, 18 tax havens have cooperated with EU countries in combattingtax evasion under the Savings Directive. When a bank in Jersey, for instance,pays interest to a French resident, it withholds 35 percent of the interest paymentas a tax and remits 75 percent of the proceeds to France without disclosing theidentity of the taxpayer. A number of havens, however, do not participate inthe Directive, most notably Singapore, Hong Kong, the Bahamas, and Bahrain.Strikingly, we find that deposit shifting in response to treaties only occurs tothe benefit of the havens that do not participate in the EU Savings Directive.As shown in col. (3), an additional treaty signed by France does not affect thedeposits in havens that apply the Directive (i.e., when STD=1), but it increasesdeposits by 1.8 percent in havens that do not apply it and do not have a treatywith France. To put it simply, deposits go to the least compliant havens. Table 2also confirms the finding of existing studies that the Directive itself significantlyaffected the bank deposits of EU residents in participating havens (Johannesen,2010).

The number of treaties signed is a crude measure of treaty coverage. Treatieswith Switzerland and Luxembourg are much more important for France in fightingtax evasion than treaties with Vanuatu and Saint Lucia. We therefore constructa second measure of treaty coverage that weighs treaties according to their im-portance.

For each country i and haven j for which we have bilateral deposit data, wecompute the share of i’s deposits in tax havens which were placed in j duringthe first year of our sample. In 2004, the location of deposits was unaffectedby the European Savings Directive which was not yet implemented, and largelyunaffected by treaties which were still few in numbers. The shares, therefore,measure the relative importance of haven j to tax evaders of country i and areexogenous to recent policy developments. For each country-haven pair (i, j), weuse the shares to weigh each treaty concluded by i with havens other than j. Theresulting measure of treaty coverage takes values between zero (no treaty) andone (full coverage). By construction, this measure only takes into account treatycoverage over the 13 havens for which we have bilateral deposit data.

As col. (4) to (6) show, with this measure of treaty coverage the results aresimilar to those obtained with the measure that merely counts the number oftreaties signed. Consider a treaty between France and a haven which, in 2004,attracted 10 percent of the deposits owned by French residents in tax havens.According to col. (4), such a treaty causes a 1.2 percent average increase in Frenchdeposits in each other BIS-reporting tax haven. As col. (5) and (6) suggest, onlythe havens that have no treaty with France and that are not covered by the EUSavings Directive attract deposits.

The results in Tables 1 and 2 show that there is a strong correlation betweentreaty signature and subsequent deposit growth in tax havens. To conclude thatthe changes in deposits we observe are caused by treaties, we need to assumethat in a counterfactual world without treaties, the deposits in the “treaty” and


“no treaty” pairs would have grown similarly. This key identifying assumptiondeserves a careful examination.

C. Tests of Identification Strategy

We have conducted two tests of our identification strategy. A first test ex-amines the possibility that tax havens might have systematically signed treatieswith countries that were placing less and less deposits in their banks relative tothe global trend, which would introduce a spurious relationship between treatysignature and deposit growth. We investigate this possibility by running probitmodels of the form:

(2) Treatyijq = α+ β2Growthijq + δXijq + γDistanceij + ζi + θq + εijq

where Treatyijq is a dummy equal to 1 if i and j sign an information exchangetreaty in quarter q, Growthijq captures the growth rate of the deposits held bysavers of country i in haven j before quarter q, Xijq includes other bilateralfactors, ζi denotes saver-country fixed effects and θq time fixed effects.

We want to know whether the probability to sign a treaty is affected by pastdeposit growth rates, i.e. whether β2 is different from zero.16 We consider twomeasures of deposit growth: the percentage growth over the 4 quarters before q,and the percentage growth from 8 quarters to 4 quarters before q. The resultsare in Table 3. As col. (1) shows, the probability to sign a treaty is not affectedby the growth rate of deposits during the year preceding treaty signature. It ismarginally affected by deposit growth from 8 quarters to 4 quarters before treatysignature, but this barely significant correlation disappears when we control fortime fixed effects (col. 2): it reflects the fact that most treaties were signed duringthe financial crisis, when deposits were falling worldwide.

Col. (3) and (4) show that the level of deposits, distance, and GDP are signif-icant determinants of the probability to sign a treaty. But when we control forthose factors, the probability to sign a treaty remains unaffected by past growthrates of deposits. On average, treaties were not concluded by country-haven pairswhere deposits were growing more slowly than the global trend.

Our second test examines whether the country-haven pairs that signed a treatyand those that did not experienced an otherwise similar evolution over the periodof study. The goal of this test is to make sure that the correlation we observebetween treaty signature and subsequent deposit growth is not driven by an un-observed third factor such as a slowdown in the financial activity of relativelycompliant havens.

16The determinants of treaty signature have been studied theoretically by Bacchetta and Espinosa(2000), Eggert and Kolmar (2002), and Huizinga and Nielsen (2003), and empirically by Ligthart,Vlachaki and Voget (2011), Bilicka and Fuest (forthcoming), and Elsayyad (2012).


Table 3—Probit Models of Treaty Signature

(1) (2) (3) (4)BANK: havens BANK: havens BANK: havens BANK: havens

VARIABLES SAVER: non-havens SAVER: non-havens SAVER: non-havens SAVER: non-havens

Deposit growth rate, -4q to 0q 0.0004 0.0011 -0.0010 -0.0013(0.6916) (0.4146) (0.6283) (0.7340)

Deposit growth rate, -8q to -4q -0.0017* -0.0012 -0.0019 -0.0037(0.0849) (0.3985) (0.2841) (0.2745)

Deposits (in logs) 0.0010** 0.0034***(0.0398) (0.0002)

Distance (in logs) -0.0041*** -0.0039*(0.0000) (0.0513)

GDP (in logs) 0.0041*** 0.0991***(0.0000) (0.0041)

Observations 56,069 37,053 11,844 4,743Time fixed effect NO YES YES YESSaver-country fixed effect NO NO NO YES

Note: p-values in parentheses, based on robust standard errors. *** denotes significance at the 1 percentthreshold, ** at the 5 percent threshold, and * at the 10 percent threshold. This table investigateswhat determines the signature of a treaty between a country i and a tax haven j. The dependentvariable is a dummy equal to 1 if a country i and haven j sign an information exchange treaty inquarter q. The unit of observation is the country-haven pair (i, j) and the sample period goes from2003q4 to 2011q2. The estimates are marginal effects. Deposit growth rate captures the growth rateof the deposits held by savers of country i in haven j before quarter q. We consider two measures ofthe growth rate of deposits: the percentage growth over the 4 quarters before q and the percentagegrowth from 8 quarters to 4 quarters before q. Deposits is the log of the stocks of deposits held bycountry i in haven j in quarter q, GDP the log of country’s i GDP (from the World Bank’s WorldDevelopment Indicator), Distance the geodesic distance between i and j (from the CEPII database,http://www.cepii.fr/anglaisgraph/bdd/distances.htm)Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.

The idea of the test is simple: if a confounding trend were driving our results,then treaty signature should be associated with a subsequent lower growth ofthe haven activities that are unrelated to treaties. So we study how those unre-lated activities evolve in the “treaty” and “no treaty” groups. We focus on theinter-bank activities of tax havens. Haven-based banks receive large amounts ofdeposits from foreign banks, which they use in turn to grant loans. Interbank de-posits received by tax havens are unrelated to personal tax evasion, so they shouldnot be affected by information exchange agreements. But they are sensitive tothe international business cycle, to domestic conditions in the havens, and moregenerally to any trend that could potentially confound our analysis of treaties. Incol. (1)-(2) of Table 4, we run the same regression for interbank deposits as wedid for the deposits owned by “non-banks” in col. (2) of Table 1 and col. (2) ofTable 2, our core specifications. The results show that treaties have zero effect oninterbank deposits. In other words, interbank deposits have evolved similarly inthe “treaty” and “no-treaty” pairs. The statistically significant effect of treatieson “non-bank” deposits is thus unlikely to be driven by an omitted differentialtime trend.

Our two tests establish that we have a reasonably valid natural experiment: the

http://www.cepii.fr/anglaisgraph/bdd/distances.htm


country-haven pairs in our sample have similar ex ante and ex post observablecharacteristics, the sole relevant difference being that some pairs signed an infor-mation exchange agreements while others did not. The correlations we documentbetween treaty signature and subsequent deposit growth can thus be consideredcausal. We present below further robustness checks.

D. Robustness Tests

OECD countries have concluded many more treaties than developing countries.Our results, one could fear, might be driven by asymmetric shocks reducing thedeposits of developed countries relative to those of developing countries, such asthe 2008-2009 financial crisis. To address this concern, we restrict the sample toOECD countries only. Col. (3)-(4) of Table 4 show that the response to treaties isslightly larger in the OECD sample than in the full sample, though qualitativelysimilar.

Second, we run the regressions with exchange rate adjusted deposit stocks. Sofar, we have used data that convert deposits in pounds, euros or Swiss francs intoU.S. dollars using end of quarter exchange rates. If a large share of bank depositsin Switzerland are denominated in Swiss francs and if Switzerland signed mostof its treaties during a period when the Swiss franc depreciated, there is a riskthat we capture a spurious effect of treaties on deposits. To address this issue, weconstruct an exchange rate adjusted measure of deposit stocks. For each country-pair, we know what fraction of deposits are denominated in U.S. dollars, euros,British pounds, Swiss francs, and yen. We use this currency decomposition tohold exchange rates fixed at their end-of-2003 level. The results are reported incol. (5)-(6) of Table 4. The estimated effects of treaties are slightly smaller butqualitatively identical to the core specifications.

This result may come as a surprise given the large exchange rate movementsduring the financial crisis. But it can easily be explained. The Online Appendixshows that the currency composition of deposits is strikingly similar in the groupof “treaty” and “no treaty” country-pairs: it is not correlated with treaty signa-ture. For this reason, exchange rate changes are absorbed by our time fixed-effectsand do not interfere with the identification of the impact of treaties.

In a final robustness check, we sequentially add country-year dummies andhaven-year dummies to the core specifications. Country-year dummies controlfor all time-varying factors at the country level, such as changes in complianceefforts, capital tax rates or the incomes of top earners who are most likely to holdassets in tax havens. Haven-year dummies control for all time-varying factorsat the haven level, such as bank crises or changes in political environment. Theresults are reported in col. (7)-(10) of Table 4. The estimated effects are robustto the inclusion of country-year dummies. When we include both country-yeardummies and haven-year dummies, we still find a modest effect of treaties ondeposits but are unable to identify a deposit shifting effect.

22

AM

ER

ICA

NEC

ON

OM

ICJO

UR

NA

LM

ON

TH

YEA

R

Table 4—Tests of Identification Strategy and of Robustness

(1) (2) (3) (4) (5) (6) (7) (8) [9] [10]

VARIABLES

Signed -0.0248 -0.0425 -0.1905*** -0.1230 -0.0890* -0.0431 -0.2962*** -0.1407* -0.1163* -0.0984(0.7963) (0.7083) (0.0094) (0.1321) (0.0954) (0.4898) (0.0001) (0.0862) (0.0744) (0.2175)

STD -0.0224 -0.5302*** -0.2279*** -0.6431*** -0.3727**(0.8235) (0.0000) (0.0002) (0.0005) (0.0211)

Treaty coverage × Signed 0.0004 0.0052 0.0015 0.0022 0.0030(0.9449) (0.1956) (0.5938) (0.6543) (0.5400)

Treaty coverage × (1- Signed) -0.0034 0.0128** 0.0125*** 0.0115** 0.0040(0.6904) (0.0210) (0.0023) (0.0151) (0.3838)

Constant 3.7524*** 3.7532*** 4.8144*** 4.7834*** 3.2197*** 3.2197*** 3.2197*** 3.2197*** 3.2197*** 3.2197***(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

Observations 20,489 20,489 8,049 8,049 30,693 30,693 8,049 8,049 8,049 8,049R-squared 0.0394 0.0395 0.0852 0.1129 0.0644 0.0693 0.1744 0.1903 0.2910 0.2941Number of panelid 1,004 1,004 307 307 1,270 1,270 307 307 307 307Countrypair fixed effects YES YES YES YES YES YES YES YES YES YESTime fixed effects YES YES YES YES YES YES YES YES YES YESSaver-year dummies NO NO NO NO NO NO YES YES YES YESBank-year dummies NO NO NO NO NO NO NO NO YES YES

SAVER: OECD SAVER: non-havens SAVER: OECDSAVER: non-havens

OECD countries only Exchange-rate adjusted Country-year fixed effectsInterbank depositsBANK: havens BANK: havens BANK: havensBANK: havens

Note: p-values in parentheses, based on robust standard errors clustered at the country-pair level. *** denotes significance at the 1 percent threshold,** at the 5 percent threshold, and * at the 10 percent threshold. The dependent variable is the stock of deposits held by savers of country i in banks oftax haven j at the end of quarter q. The unit of observation is the country-haven pair (i, j) and the sample period goes from 2003q4 to 2011q2. Signedis a dummy equal to 1 if there exists a treaty providing for information exchange between i and j in quarter q. STD is a dummy equal to one if thecountry-haven pair (i, j) applies the EU Savings Directive. Treaty coverage counts the number of treaties that i has with tax havens other than j. Col.(3)-(10) consider the deposits held by non-bank agents; col. (1)-(2) the deposits held by banks.Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.


IV. Deposits Held Through Sham Corporations

There is a great deal of anecdotal evidence suggesting that clients of offshorebanks routinely use sham corporations with addresses in tax havens such asPanama as nominal owners of their bank accounts in Switzerland and otherhavens. The IRS, for instance, provides case studies of tax evasion by U.S. indi-viduals through a big Swiss bank revealing a quasi-systematic use of shell com-panies.17 This section focuses on how deposits held through sham corporationshave responded to the wave of tax treaties.

Remember that when a French saver holds assets in Switzerland through a shamPanamanian company, the BIS assigns the funds to Panama. This conventionexplains why haven-haven deposits are so important in the BIS statistics: in thefirst half of 2011, they accounted for around $550 billion, almost 25 percent of allthe deposits in tax havens. Deposits from the British Virgin Islands and Panamawere particularly important. Both jurisdictions have flexible corporate laws thatmake it simple to create companies in a few minutes.

Using a sham corporation as nominal account holder adds a layer of secrecybetween an account and its beneficial owner: essentially, accounts held throughsham corporations are equivalent to numbered accounts, which are today prohib-ited by anti-money laundering regulations. Sham corporations also help avoidingtaxes: the EU Savings Directive does not apply to the deposits held by Europeanresidents through sham companies. But they do not protect from informationexchange treaties. If France and Switzerland have a treaty and French authori-ties suspect a taxpayer of hiding funds in Switzerland, they can ask Switzerlandto provide the relevant information, even if the funds are held through a shellcompany. Banks are required by anti-money laundering regulations to know atall times who are the ultimate owner of the assets they manage. They must pro-vide this information to foreign authorities that file information requests under atreaty.

The implication is that if tax evaders respond to treaty signature, then treatiesconcluded between havens like Switzerland and countries like France should affectthe Swiss deposits held by French residents through sham corporations, i.e. theSwiss deposits that the BIS assigns to the British Virgin Islands, Panama, andother havens.

Table 5 investigates whether this is the case by analyzing the evolution of haven-haven deposits. In col. (1), we regress haven-haven deposits (e.g., Swiss depositsassigned to Panama) on the number of treaties concluded by banking havens (e.g.,Switzerland) with non-haven countries (e.g., France). A treaty between Franceand Switzerland reduces the Swiss deposits registered as belonging to each taxhaven by 0.7 percent on average.

17See http://www.irs.gov/uac/Offshore-Tax-Avoidance-and-IRS-Compliance-Efforts. See alsoZaki (2010) for anecdotal evidence on the use of sham corporations by Europeans, and Hanlon, Maydewand Thornock (2011) for evidence on the use of sham offshore corporations by U.S. tax evaders for theirU.S. investments.

http://www.irs.gov/uac/Offshore-Tax-Avoidance-and-IRS-Compliance-Efforts


Table 5—Panel Regressions of Bank Deposits Held Through Sham Corporations

(1) (2) (3) (4)BANK: havens BANK: havens BANK: havens BANK: havens

SAVER: havens SAVER: havens SAVER: havens SAVER: havens

VARIABLESTREATY COVERAGE:




share

-0.0067** -0.0095*** -0.5900*** -0.6045***(0.0188) (0.0015) (0.0000) (0.0000)

0.0087 0.0224(0.3362) (0.9103)

0.0536 0.1005(0.6726) (0.4022)

4.3572*** 4.3604*** 4.4043*** 4.4057***(0.0000) (0.0000) (0.0000) (0.0000)

Observations 8,798 8,798 8,798 8,798R-squared 0.1188 0.1199 0.1359 0.1365Number of panelid 346 346 346 346Countrypair fixed effect YES YES YES YESTime fixed effect YES YES YES YES

Treaty coverage, banking haven with non-haven countries

Treaty coverage, banking haven with other tax havens

Signed

Constant

Note: p-values in parentheses, based on robust standard errors clustered at the country-pair level. ***denotes significance at the 1 percent threshold, ** at the 5 percent threshold, and * at the 10 percentthreshold. The table investigates how the signature of a treaty between a tax haven (e.g., Switzerland)and a non-haven country (e.g., France) affects the deposits recorded by the BIS as belonging to taxhavens (e.g., the deposits in Swiss banks recorded as belonging to Panama). The dependent variableis the stock of deposits recorded as belonging to haven i (e.g., Panama) in the banks of haven j (e.g.,Switzerland) at the end of quarter q. The unit of observation is the haven-haven pair (i, j) and the sampleperiod goes from 2003q4 to 2011q2. For a given banking haven j, there are up to 41 “saving” havensi. We consider the deposits held in 13 banking havens j. In col. (1)-(2), Treaty coverage, bankinghaven with non-havens counts the number of treaties that j has with non-haven countries (and Treatycoverage,banking haven with other tax havens the number of treaties that j has with other havens). Incol. (3)-(4), the Treaty coverage, variables measure the share of the deposits held by non-haven (reps.haven) countries in haven j in 2004 that are covered by a treaty in quarter q. Signed is a dummy equalto 1 if there exists a treaty providing for information exchange between haven i and haven j in quarterq.Source: Bank for International Settlements (2002-2011), restricted bilateral locational banking statistics.

In col. (2), we investigate whether haven-haven treaties matter for the pat-tern of haven-haven deposits. Neither a treaty between Switzerland and Panama(Signed = 1) nor treaties between Switzerland and havens other than Panamaaffect the value of the Swiss deposits assigned to Panama in the BIS statistics,which is fully consistent with our interpretation of what haven-haven depositsrepresent. Indeed, there is no reason why information exchange between Panamaand Switzerland should affect the French residents who use sham corporations inPanama as nominal owners of their Swiss accounts.

In col. (3) and (4), we run the same regressions as in col. (1) and (2) butwith the measure of treaty coverage that weighs treaties by the importance ofthe deposits covered. The estimated effects are statistically and economicallysignificant. Consider a treaty between France and Switzerland. Assume thatFrench residents hold 10 percent of all Swiss deposits belonging to non-havencountries. Col. (3) suggests that such a treaty reduces the bank deposits in


Switzerland registered as belonging to tax havens (e.g., Panama) by 4.5 percent.18

Now assume that French residents are also the ultimate owners of 10 percent ofthe Swiss deposits registered as belonging to tax havens. Under this assumption,a treaty between France and Switzerland causes a 45 percent reduction of thedeposits held in Switzerland by French savers through sham corporations. Underplausible assumptions, the tax evaders who use sham corporations may haveresponded strongly to the G20 crackdown.

There is one caveat, however: since we cannot identify the ultimate ownersof the deposits held through sham corporations, the results in Table 5 rely onvariation at the haven level rather than variation at the country-haven-pair level.It is an unfortunate feature of cross-border bank deposits statistics that theyare based on immediate rather than beneficial ownership. If deposit data wereestablished on a beneficial ownership basis, almost no deposits would be assignedto the British Virgin Islands or Panama; more deposits would be assigned to theU.S., Italy, or France; and it would be easier to track the progress made in thefight against tax evasion.

V. The Compliance Effect of Treaties

Our results so far indicate that the G20 initiative has caused a relocation ofdeposits between tax havens leaving the funds globally held offshore roughly un-changed. But depositors may have responded to the crackdown by complyingmore with tax laws while keeping their funds in tax havens. In this section weanalyze the available evidence on the compliance effect of treaties.

There are two types of data at hand. First, we have direct information ontax compliance in Switzerland, probably the most important tax haven as far aspersonal wealth management is concerned.19 Since mid-2005, in the context ofthe EU Savings Directive, Swiss banks must withhold a tax on interest incomepaid to European households who own Swiss accounts. Savers can escape thewithholding tax if they voluntarily declare their income to their home countrytax authority. Swiss authorities have published on a yearly basis the amount ofinterest earned by residents of each EU country, as well as what fraction of thisincome savers have chosen to voluntarily disclose. We know for instance that in2011, French residents earned CHF 324 million in interest, and chose to declare33 million, or about 10 percent. To our knowledge, this unique dataset has neverbeen used before in the literature.20

It enables us, for one key haven and 27 counterpart countries, to conduct adirect test of the compliance effect of treaties. We analyze how the share ofinterest declared has evolved over 2006-2011 for the 15 EU countries that have

18(exp(−0.59) − 1) × 10 = 4.5 percent.19Switzerland comes second to the Cayman Islands in terms of deposits, but an exceptionally high

fraction of deposits in Swiss banks seem to belong to individuals (80-90 percent, whereas our informedguess for the average across all havens is about 50 percent).

20The data are available on the authors’ websites.


signed a treaty with Switzerland since 2008 (e.g., France, Spain, Austria), andfor the 12 countries that have not (e.g., Belgium, Portugal, Hungary). As shownby Figure 7, there has been a general increase in compliance over the 2006-2011period. But there is no indication that this trend has been any stronger for thecountries that have signed a treaty with Switzerland. And indeed, when we usethe same regression framework as in Section III, we find that treaty signature hasno statistically significant effect on the fraction of interest that taxpayers choseto declare.21 Despite the G20 initiative, the general level of compliance of EUSwiss bank account holders remains low, around 10-20 percent.22

-4%

1%

6%

11%

16%

21%

0%

5%

10%

15%

20%

25%

2006 2007 2008 2009 2010 2011

Sha

re o

f int

eres

t inc

ome

earn

ed in

Sw

itzer

land

de

clar

ed to

hom

e co

untr

y ta

x au

thor

ities

Residents of EU countries that signed a treaty with Switzerland (left-hand scale)

Residents of EU countries that didn't sign a treaty with Switerland (right-hand scale)

Figure 7. Fraction of Interest Income Earned by EU Residents in Swiss Banks Declared to

Home Country Tax Authorities

Source: Administration federale des contributions.

The second type of evidence on tax enforcement comes from the OECD (2011),which has gathered data on the amount of taxes recovered due to increased com-pliance on the part of offshore account holders. Over the 2009-2011 period, theOECD (2011) reports an increase of almost EUR 14 billion in taxes paid in richcountries. This is certainly far from negligible. However, assuming that evaders

21See Online Appendix.22The compliance figures reported on Figure 7 are upper bounds, for one simple reason. They are

obtained by dividing interest declared by interest earned, but the denominator excludes interest earnedby EU residents through sham corporations, and a very large fraction of Swiss bank fiduciary depositsare held through sham corporations.


paid in taxes and penalties an amount equivalent to 5 percent of their assets(which is what the OECD reports for Italy, Mexico, and the UK), then the OECDfigures imply that about $350bn in offshore assets may have been disclosed to taxauthorities. This figure falls short of the $6,000bn or so likely held by householdsin tax havens.23 Taken at face value, the OECD’s findings do not lend supportto the view that compliance has considerably improved.

The evidence we have just described is far from systematic. There is no cross-country database on tax compliance comparable to the BIS’ bank deposit statis-tics. So we cannot fully exclude a large increase in compliance in havens otherthan Switzerland. Better measuring compliance and its determinants is an im-portant challenge for future research.

VI. Concluding Remarks

Conventional wisdom among policymakers is that the G20 tax haven crackdownis a success. The evidence presented in this paper challenges this view. It suggeststhat, so far, treaties have led to a relocation of bank deposits between tax havensbut have not triggered significant repatriations of funds. The least complianthavens have attracted new clients, while the most compliant ones have lost some,leaving roughly unchanged the total amount of wealth managed offshore.

Although this is disappointing, we emphasize that the G20 initiative is notuseless. We find evidence that some tax evaders have responded to the wave oftax treaties. Many experts were skeptical that upon request information sharingcould achieve anything at all. Our results belie the most pessimistic views onthe efficacy of treaties: even a weak threat of enforcement is sometimes enoughto affect behavior. Further, uncertainties remain on the extent to which treatieshave induced tax evaders to comply more with tax laws while keeping their fundsoffshore.

Yet our results suggest that there is room to improve the fight against offshoretax evasion. First, the G20 could urge tax havens to sign treaties with all coun-tries: a comprehensive multilateral agreement would prevent tax evaders fromtransferring their funds from haven to haven. Second, our results suggest thateven in the presence of a complete network of upon request information exchangetreaties, there may remain a scope for improved tax collection by making treatiesmore demanding.

The G20 tax haven crackdown is a major coordinated initiative against taxevasion at the global level. Another important initiative, at the regional level, isthe European Union Savings Directive. The G20 initiative relies on informationexchange treaties; the EU Savings Directive imposes a withholding tax on interestincome earned by European residents in a number of cooperating tax havens. So

23Based on interviews with wealth managers, the Boston Consulting Group (2010) puts the amountof offshore wealth at $7,400bn in 2009. This figure is close to the one found by Zucman (2013), whoreckons that 8 percent of households’ financial wealth is held in tax havens, which is around $6,000bn in2008.


far, both policies have pitfalls: treaties are not comprehensive enough; the EUwithholding tax exempts equities and derivatives, and does not look through shamcorporations that tax evaders routinely use (Johannesen, 2010; Zucman, 2013).Therefore, what is the best tool treaty or tax to combat offshore tax evasionremains an open question.

A comprehensive network of treaties providing for automatic exchange of infor-mation would put an end to bank secrecy and could make tax evasion impossible.Taxes withheld on all incomes earned by foreign residents in all tax havens couldalso make tax evasion impossible, while maintaing some form of bank secrecy.Which of the two instruments would maximize tax revenues while minimizingadministrative costs, including the costs of negotiating with tax havens? There isneed for more research on this question. Policymakers have diverging views: onthe one hand, the European Union Commission pushes for automatic exchangeof information, just like the U.S. with the Foreign Account Tax Compliance Act(FATCA), but on the other hand countries such as Germany and the U.K. arenegotiating a comprehensive withholding tax with Switzerland.

Another question raised by our study is why some havens cooperate more thanothers. Tax havens have a strong economic interest in bank secrecy. But maybeabandoning bank secrecy has a positive effect on a haven’s reputation, which mayhelp it attract other financial activity, such as the incorporation of investmentfunds. This issue would deserve to be further analyzed.

REFERENCES

Alworth, Julian S., and S. Andresen. 1992. “The Determinants of Cross-Border Non-Bank Deposits and the Competitivness of Financial Market Cen-tres.” Money Affairs, 5: 105–133.

Bacchetta, Philippe, and Maria Paz Espinosa. 2000. “Exchange-of-Information Clauses in International Tax Treaties.” International Tax and Pub-lic Finance, 7: 275–293.

Bank for International Settlements. 2002-2011. “Locational Banking Statis-tics: 3. External Loans and Deposits of banks in Individual ReportingCountries.” http: // www. bis. org/ statistics/ bankstats. htm (accessedNovember 27, 2011).

Bank for International Settlements. 2006. “Guidelines to the internationallocational banking statistics.” December 2008 ed., Bank for International Set-tlements.

Bilicka, Katarzyna, and Clemens Fuest. forthcoming. “With which countriesdo tax havens share information?” International Tax and Public Finance.

Blonigen, Bruce A., and Ronald B. Davies. 2005. “Do Bilateral Tax TreatiesPromote Foreign Direct Investment?” In Handbook of International Trade,

http://www.bis.org/statistics/bankstats.htm


Volume II: Economic and Legal Analysis of Laws and Institutions. , ed. J.Hartigan, 526–546. Blackwell Publishers.

Blonigen, Bruce A., Lindsay Oldenski, and Nicholas Sly. 2011. “Separat-ing the Opposing Effect of Bilateral Tax Treaties.” NBER Working Paper.

Boston Consulting Group. 2010. “Global Wealth Report, Regaining LostGround.”

Brown, Martin, Barbara Dobeli, and Philip Saure. 2011. “Swiss BankAccounts.” Working Paper.

Chan, Kalok, Vicentiu M. Covrig, and Lilian K. Ng. 2005. “What De-termines the Domestic Bias and Foreign Bias? Evidence from Mutual FundEquity Allocations Worldwide.” The Journal of Finance, 60(3): 1495–1534.

Chetty, Raj, Adam Looney, and Kory Kroft. 2009. “Salience and Taxation:Theory and Evidence.” American Economic Review, 99(4): 1145–1177.

Desai, Mihir A., and Dhammika Dharmapala. 2011. “Dividend Taxesand International Portfolio Choice.” The Review of Economics and Statistics,93: 266–284.

Desai, Mihir A., C. Fritz Foley, and James R. Hines. 2006. “The Demandfor Tax Haven Operations.” Journal of Public Economics, 90(3): 515–531.

Dharmapala, Dhammika. 2008. “What Problems and Opportunities are Cre-ated by Tax Havens?” Oxford Review of Economic Policy, 24: 661–679.

Dharmapala, Dhammika, and James R. Hines. 2009. “Which CountriesBecome Tax Havens?” Journal of Public Economics, 93: 1058–1068.

di Giovanni, Julian. 2005. “What Drives Capital Flows? The Case of Cross-Border M&A Activity and Financial Deepening.” Journal of International Eco-nomics, 65(1): 127–149.

Eggert, Wolfgang, and Martin Kolmar. 2002. “Residence-Based CapitalTaxation in a Small Open Economy: Why Information is Voluntarily Ex-changed and Why it is Not.” International Tax and Public Finance, 9(4): 465–482.

Elsayyad, May. 2012. “Bargaining over tax information exchange.” Max PlanckInstitute for Tax Law and Public Finance Working Paper.

Elsayyad, May, and Kai A. Konrad. 2011. “Fighting Multiple Tax Havens.”Journal of International Economics, 86(2): 295–305.

G20. 2009. “The Global Plan for Recovery and Reform, Final Communique ofthe G20 Summit Held in London on 2 April 2009.” Available online at http:// www. g20. org/ documents/ final-communique. pdf .

http://www.g20.org/documents/final-communique.pdf

http://www.g20.org/documents/final-communique.pdf


Global Forum on Transparency and Exchange of Information for TaxPurposes. 1998-2011. “Exchange of Information Portal.” http: // eoi-tax.

org/ (accessed November 12, 2011).

Hanlon, Michelle, Edward L. Maydew, and Jacob R. Thornock. 2011.“Taking the Long Way Home: Offshore Investments in U.S. Equity and DebtMarkets and U.S. Tax Evasion.” Working Paper.

Hemmelgarn, Thomas, and Gaetan Nicodeme. 2009. “Tax Co-ordinationin Europe: Assessing the First Years of the EU-Savings Taxation Directive.”European Commission Working Paper.

Hines, James R., and Eric M. Rice. 1994. “Fiscal Paradise: Foreign TaxHavens and American Business.” Quarterly Journal of Economics, 109(1): 149–182.

Huizinga, Harry, and Gaetan Nicodeme. 2004. “Are international depositstax-driven?” Journal of Public Economics, 88: 1093–118.

Huizinga, Harry, and Soeren Bo Nielsen. 2003. “Withholding Taxes orInformation Exchange: the Taxation of International Interest Flows.” Journalof Public Economics, 87: 39–72.

Johannesen, Niels. 2010. “Tax Evasion and Swiss Bank Deposits.” EPRUWorking Paper 2010-05.

Keen, Michael, and Jenny E. Ligthart. 2006. “Information Sharing andInternational Taxation: a Primer.” International Tax and Public Finance,13(1): 81–110.

Kleven, Henrik, Martin Knudsen, Claus Kreiner, Soren Pedersen, andEmmanuel Saez. 2011. “Unwilling or Unable to Cheat? Evidence from a TaxAudit Experiment in Denmark.” Econometrica, 79(3): 651–692.

Ligthart, Jenny E., Mina Vlachaki, and Johannes Voget. 2011. “TheDeterminants of Double Tax Treaty Formation.” mimeo.

Louie, Henry J., and Donald J. Rousslang. 2008. “Host-Country Gover-nance, Tax Treaties and U.S. Direct Investment Abroad.” International Taxand Public Finance, 15(3): 256–273.

OECD. 2008. “Model Tax Convention on Income and on Capital.” OECD Com-mitte on Fiscal Affairs.

OECD. 2011. “The Era of Bank Secrecy is Over.” The G20/OECD process isdelivering results, available online at http: // www. oecd. org/ dataoecd/ 54/60/ 48996146. pdf .

http://eoi-tax.org/

http://eoi-tax.org/

http://www.oecd.org/dataoecd/54/60/48996146.pdf

http://www.oecd.org/dataoecd/54/60/48996146.pdf


Palan, Ronen, Richard Murphy, and Christian Chavagneux. 2010. TaxHavens: How Globalization Really Works. Cornell Studies in Money, Ithaca,NY:Cornell University Press.

Shaxson, Nicholas, and John Christensen. 2011. “Time to Black-List theTax Haven Whitewash.” Financial Times, April 4th.

Sheppard, Lee. 2009. “Don’t Ask, Don’t Tell, Part 4: Ineffectual InformationSharing.” Tax Notes.

U.S. Government Accountability Office. 2011. “IRS’s Information Ex-changes with Other Countries Could Be Improved through Better PerformanceInformation.” Report to the Permanent Subcommittee on Investigations, U.S.Senate.

U.S. Senate. 2008. “Tax Haven Banks and U.S. Tax Compliance.” Staff Reportof the Permanent Subcommitee on Investigations.

Zaki, Miret. 2010. Le secret bancaire est mort, vive l’evasion fiscale ! Lau-sanne:Favre.

Zucman, Gabriel. 2013. “The Missing Wealth of Nations: Are Europe and theU.S. net Debtors or net Creditors?” Quarterly Journal of Economics, forth-coming.

Documents

Niels Johannesen and Gabriel Zucman The end of bank ...eprints.lse.ac.uk/56125/1/__lse.ac.uk_storage... · bank secrecy. Exploiting a unique panel dataset, our study is the rst attempt